Public health insurance, individual health, and entry into self-employment

Even in countries not following Germany’s “Bismarck”-type model, public and private health insurance often co-exist. For example, Besley et al. (1998, 1999) analyze the interactions between the public and private health care sectors in the UK.

We use the term “entrepreneurship lock” as it is used in the literature, describing a barrier to entry into entrepreneurship (Fairlie et al. 2011), i.e., a lockout effect. In contrast, the original term “job lock” means that employees are locked into their current jobs.

Gumus and Regan (2015) find a statistically significant effect on entry only for singles and married men whose wives lack employer-provided health insurance.

A related, but distinct stream of literature investigates the determinants of health insurance coverage of the self-employed. This issue is of special interest in the USA, where a large number of self-employed persons lack health insurance (Perry and Rosen 2004). A number of articles (Gruber and Poterba 1994; Heim and Lurie 2009; Selden 2009; Gumus and Regan 2013) estimate the impact of the improved tax deductions of health insurance premiums brought by TRA86 on health insurance demand among the self-employed. They all find significant effects in the expected direction.

Fossen (2009) simulates individual net income after taxes, all social insurance contributions, and transfers in Germany and focuses on the effects of personal income tax reforms on transitions into and out of self-employment.

According to numbers reported by Gress et al. (2005), only 0.12% of the paid employees and 0.85% of the self-employed were not covered under any health insurance in 2003.

Until 2002, the obligatory insurance limit and the contribution assessment ceiling were identical.

These provisions are intended to allow for some degree of competition between the SHI funds, but in practice, competition is rather limited because services are regulated to a large extend and therefore very similar. The differences in contribution rates were rather small, and the extra premium was at most €8 per month. Table A 1 in Appendix A shows the contribution rates of SHI funds and groups of SHI funds by year. Historically, most SHI funds originate from occupational groups or regions, and many insurees simply stay with their parents’ SHI fund. The introduction of the more salient extra premiums may have triggered more awareness and more switching between SHI funds since 2009 (Schmitz and Ziebarth 2016).

PHI also includes private long-term care insurance.

In Saxony, employees pay one percentage point of the long-term care insurance contributions alone, to compensate for an additional holiday.

Exceptions apply for artists, writers, and journalists.

In general, this lower limit is three quarters of a defined reference value, which is adjusted annually. It rose from €2291 per month in 2000 to €2625 in 2011.

We discuss potentially different qualities of healthcare and possible shifting of costs further below.

In Sect. 6.2, we show that the estimated effects do not change if we assume normally distributed error terms instead, which leads to a probit model.

The HICD amount depends on the assumption about cost shifting from employers to employees. As mentioned before, in the baseline estimations, we assume that the perceived economic incidence equals the statutory incidence, i.e., the burden is roughly split by half. We assess the sensitivity of the results with respect to this assumption in a robustness check (see Sect. 6.2).

With full contribution rate, which enables the receipt of sickness benefits.

One might be concerned about potential endogeneity of this control variable. However, its coefficient turns out to be insignificant in all our estimations, and dropping this control variable does not change the other coefficients notably.

As an example for the calculation of the HICD, consider a family forming a household with a male main earner in full-time paid employment and covered under SHI, a wife not participating in the labor market, who stays at home to take care of two young children, and a young adult daughter in paid employment with SHI. The wife and the two children are covered contribution-free under family insurance. If the main earner switches to self-employment, he can stay in voluntary SHI but has to pay both the employee’s and employer’s shares of the contributions. If he switches to PHI, his wife and the young children lose their family insurance, and the family must buy PHI plans for them as well. The adult daughter remains covered under her SHI. We assume that the cheaper alternative of voluntary SHI or PHI for the whole family is chosen. If somebody in the family is sick (except the adult daughter, who is covered by SHI in any case) or the parents are older, voluntary SHI is likely to be the cheaper option for the family. In contrast, if the adult daughter switches to self-employment, nothing changes for the father, who keeps his SHI coverage, and the mother and young children, who continue to be covered under family insurance through the father’s SHI membership. The adult daughter can choose voluntary SHI or PHI for herself without further consequences to the others. PHI will be cheaper for the young women if she is healthy and has higher income. Whatever the cheaper choice, the total HICD for the household differs between the different adult household members potentially entering self-employment, so we have variation even within households.

The effect of children on the entry probability is allowed to vary by gender by including an interaction term.

The general willingness to take risks is observed in 2004, 2006, and in all years since 2008. In the other years, we impute values observed for the same person in other years.

In a robustness check, we include nine occupation dummy variables instead of the industry dummies. The results of interest do not change significantly.

The last year is only used to determine if a transition to self-employment occurred between 2011 and 2012.

The self-employed may or may not employ workers. The concept of entrepreneurship may differ from the concept of self-employment. Entrepreneurship usually implies risk bearing and innovation, whereas self-employment goes along with income risk but not necessarily with innovation.

We obtain similar results when constraining the sample to older or younger workers.

We include civil servants in the sample in a robustness check in Sect. 6.2.

The last three rows of Table 2 show alternative measures of the real HICD under different assumptions, which we discuss in Sect. 6.2.

We obtain very similar standard errors when clustering at the household level, and the reported significance levels of the variables of interest remain unchanged.

The elasticities are derived as follows. This paper: (280)/100 (minimum health insurance costs in self-employment in units of €100, see Table 2) × 1% × (−0.0015) (coefficient of HICD) / 0.009 (mean entry rate) = −0.47%. Gumus and Regan (2015): 1% × (−0.0136) (coefficient of log tax price) / 0.0243 (entry rate in 1999) = −0.56%. Heim and Lurie (2010): 1% × (−0.117) (coefficient of log tax price) / 0.04 (mean entry rate) = −2.93%.

The estimated effect of the HICD remains virtually unchanged when we include dummy variables for each possible answer to the subjective health question instead of the continuous variable and also when we additionally control for the official degree of disability.

We thank two anonymous reviewers for suggesting some of these robustness checks to us.

For the UK, Jones and Latreille (2011) report that nonwork-limited disabled persons are less likely to be self-employed than the nondisabled, whereas work-limited disabled men are more likely to be self-employed, which could be explained by the greater flexibility self-employment offers in terms of times, hours, and locations.

We additionally include a self-employment dummy variable in the PHI premium regression and set it to one when we predict health insurance premiums for the case of self-employment.

We also obtain very similar results when we additionally include civil servants in the PHI premium regression (2007 observations in total) and in the subsequent estimation of the probability of entry into self-employment. The lower PHI contributions of civil servants are accounted for by a dummy variable indicating eligibility for civil servant healthcare.